Treasury says forecasting a single figure is too difficult: It wants confidence intervals in its forecasts

At this year’s budget announcement, the Australian Federal Treasury was 70% sure GDP would grow between 2% and 3.75%.

It might seem like a large margin of error. But that’s how Treasury wants all its forecasts to be understood.

In a paper released yesterday, the Treasury has proposed a new approach to its forecasts.

Instead of a single number, it wants policy-makers and those in the media to coach their understanding of its figures with reference to confidence intervals. Confidence intervals, a mainstay of statistical analysis, are a measure of uncertainty around a forecast. They generally do not survive the journey from economist’s report to politician’s lips.

Treasury has calculated its confidence over its own forecast by going back over how accurate they were from the period beginning 1999.

“Rather than focusing on precise point estimates, a more nuanced discussion would acknowledge that uncertainty is an unavoidable feature of forecasts, with confidence intervals spanning a wide range of outcomes,” the report, written by seven Treasury economists, states.

“Confidence intervals provide a guide to the degree of uncertainty around forecasts… This is particularly true of the nominal macroeconomic and fiscal variables, for which uncertainty increases as the forecast horizon lengthens.”

Reporting confidence intervals is something several other forecasting agencies already do, the paper notes, citing the RBA, the US Federal Reserve, and the European Central Bank.

Earlier this year, Treasury released a paper that looked at how accurate its forecasts have been in retrospect.

That paper, written by Access Economics forecaster and former Treasury official David Chessell, found that “with the benefit of hindsight, Treasury has tended to underestimate growth during economic upswings and overestimate growth during economic downturns”.

The biggest problem for Treasury, Chessell noted, was forecasting company tax revenues. Overall, the report was fairly complementary.

“Over the past two decades, Budget forecasts of nominal economic growth have exhibited a mean absolute percentage error of 1.6 percentage points,” Chessell wrote.

However, MacroBusiness founder David Llewellyn-Smith notes that the Chessel review found Treasury made larger forecasting errors than Canada, the United Kingdom and the United States.

Treasury forecasts came under fire during the recent federal election campaign, when then shadow treasurer Joe Hockey said he would not be relying on Treasury’s pre-election fiscal statement because he couldn’t trust the forecasts.